Blood and heart

Blood is very important to ensure the proper functioning of the body. Its functions cover the transport of oxygen and nutrients to the cells of the body as well as the removal of the cellular waste products.

Blood is transported to the rest of the body because it is pumped by the heart. This organ receives oxygen-poor blood and sends it to the lungs to oxygenate it. And sends the oxygen-rich blood that comes from the lugns to the rest of the body.

By josiño - Own work, Public Domain, https://commons.wikimedia.org/w/index.php?curid=9396374. Flow of the blood through the chambers of the heart. Blue arrows represent oxygen-poor blood received from the rest of the body and sent to the lungs. Red arrows represent oxygen-rich blood coming from the lungs that is sent to the rest of the body.

An inadequate supply of the blood can yield the cells to not get enough energy to function properly, causing the death of the cells in the worst case.

Coronary heart disease

The heart also needs oxygen and nutrients to function properly, these come through arteries known as coronary arteries. When we talk about a coronary disease, we often mean a difficulty of the blood flow in these arteries due to the accumulation of substances on their walls.

By NIH: National Heart, Lung and Blood Institute - http://www.nhlbi.nih.gov/health/health-topics/topics/heartattack/, Public Domain, https://commons.wikimedia.org/w/index.php?curid=25287085. Death of heart cells due to an ischemia in the coronary arteries.

In the worst case, the impact of leaving the cells of the heart without nutrients and oxygen is a heart attack, in other words, the death of part of the heart cells. This, in turn, would have an impact on the rest of the body because the pumping of the heart would be affected.

Glossary of terms

• Atherosclerosis: accumulation of substances on the walls of arteries which can hinder the blood flow. Moreover, the rupture of this plaque of substances can cause the formation of a blood clot (thrombus) that, in turn, can block even more the affected area or go to other parts of the body and block those parts (embolism). (Sources: American Heart Association, Mayo Clinic)

• Ischemia: blood flow reduction to a tissue. This implies a reduction of the supply of oxygen and nutrients, so cells won’t get enough energy to function properly. (Sources: American Heart Association, Mayo Clinic, Wikipedia)

• Angina: chest pain due to a blood flow reduction in the coronary arteries. (Sources: United Kingdom National Health Service, (Spanish) Video sobre angina de Alberto Sanagustín)

• Stable angina: angina caused by situations that requires oxygen (for example, exercise or stress) and it goes away on rest.

• Unstable angina: angina that can happen even on rest.

• Typican & atypical angina: typical angina usually means a chest disconfort. However, looks like some people can experience other symptoms like nausea or shortness of breath. In these cases people talk about atypical angina. (Sources: Harrington Hospital, Wikipedia).

• Thrombus: blood mass in solid state that hinders the blood flow in a blood vessel. (Source: MedlinePlus)

• Embolus: thrombus that detatches and goes to other parts of the body. (Source: MedlinePlus)

• Acute myocardial infarction: also known as heart attack, is the death of part of the heart tissue due to an ischemia. In other words, it is the death of part of the cells due to the lack of oxygen. (Sources: Healthline, Wikipedia)

• Electrocardiogram: graph record of the electric signals that causes heart beats. Each part of the record of a normal heart beat has a name, the most interesting ones for this project are the T wave and the ST segment because they can give some information about the presence of issues like an ischemia or infarction. (Sources: Mayo Clinic, Wikipedia, (Spanish) Video sobre electrocardiograma de Alberto Sanagustín, (Spanish) Serie de videos sobre el electrocardiograma normal de Alberto Sangaustín)

• Nuclear stress test: a radioactive dye is injected into the patient to see the blood flow on rest and doing exercise. Moreover, during this test the activity of the heart is also measured with an electrocardiogram. (Sources: Mayo Clinic, Healthline)

• Asymptomatic disease: a disease that a patient has but they experience very few or no symptoms. (Sources: (Spanish) definicion.de, MayoClinic, Wikipedia)

• Left ventricular hypertrophy: thickening of the walls of the main heart chamber that pumps the blood to the rest of the body. This can cause the muscle to loose elasticity which, in turns, causes the heart to not work properly. (Sources: Mayo Clinic)

1.1 Variable type

There are different functions in R that help us summarize the type of variables we have. The function glimpse, summary, or skim.

glimpse(data)

## Rows: 303
## Columns: 15
## $ X        <int> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18…
## $ age      <int> 63, 37, 41, 56, 57, 57, 56, 44, 52, 57, 54, 48, 49, 64, 58, 5…
## $ sex      <int> 1, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1…
## $ cp       <int> 3, 2, 1, 1, 0, 0, 1, 1, 2, 2, 0, 2, 1, 3, 3, 2, 2, 3, 0, 3, 0…
## $ trestbps <int> 145, 130, 130, 120, 120, 140, 140, 120, 172, 150, 140, 130, 1…
## $ chol     <int> 233, 250, 204, 236, 354, 192, 294, 263, 199, 168, 239, 275, 2…
## $ fbs      <int> 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0…
## $ restecg  <int> 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1…
## $ thalach  <int> 150, 187, 172, 178, 163, 148, 153, 173, 162, 174, 160, 139, 1…
## $ exang    <int> 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0…
## $ oldpeak  <dbl> 2.3, 3.5, 1.4, 0.8, 0.6, 0.4, 1.3, 0.0, 0.5, 1.6, 1.2, 0.2, 0…
## $ slope    <int> 0, 0, 2, 2, 2, 1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 1, 2, 0, 2, 2, 1…
## $ ca       <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0…
## $ thal     <int> 1, 2, 2, 2, 2, 1, 2, 3, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3…
## $ target   <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…

summary(data)

##        X              age             sex               cp       
##  Min.   :  1.0   Min.   :29.00   Min.   :0.0000   Min.   :0.000  
##  1st Qu.: 76.5   1st Qu.:47.50   1st Qu.:0.0000   1st Qu.:0.000  
##  Median :152.0   Median :55.00   Median :1.0000   Median :1.000  
##  Mean   :152.0   Mean   :54.37   Mean   :0.6832   Mean   :0.967  
##  3rd Qu.:227.5   3rd Qu.:61.00   3rd Qu.:1.0000   3rd Qu.:2.000  
##  Max.   :303.0   Max.   :77.00   Max.   :1.0000   Max.   :3.000  
##                                                                  
##     trestbps          chol            fbs            restecg      
##  Min.   : 94.0   Min.   :126.0   Min.   :0.0000   Min.   :0.0000  
##  1st Qu.:120.0   1st Qu.:211.0   1st Qu.:0.0000   1st Qu.:0.0000  
##  Median :130.0   Median :240.0   Median :0.0000   Median :1.0000  
##  Mean   :131.6   Mean   :246.3   Mean   :0.1485   Mean   :0.5281  
##  3rd Qu.:140.0   3rd Qu.:274.5   3rd Qu.:0.0000   3rd Qu.:1.0000  
##  Max.   :200.0   Max.   :564.0   Max.   :1.0000   Max.   :2.0000  
##                                                                   
##     thalach          exang           oldpeak         slope      
##  Min.   : 71.0   Min.   :0.0000   Min.   :0.00   Min.   :0.000  
##  1st Qu.:133.5   1st Qu.:0.0000   1st Qu.:0.00   1st Qu.:1.000  
##  Median :153.0   Median :0.0000   Median :0.80   Median :1.000  
##  Mean   :149.6   Mean   :0.3267   Mean   :1.04   Mean   :1.399  
##  3rd Qu.:166.0   3rd Qu.:1.0000   3rd Qu.:1.60   3rd Qu.:2.000  
##  Max.   :202.0   Max.   :1.0000   Max.   :6.20   Max.   :2.000  
##                                                                 
##        ca              thal           target      
##  Min.   :0.0000   Min.   :1.000   Min.   :0.0000  
##  1st Qu.:0.0000   1st Qu.:2.000   1st Qu.:0.0000  
##  Median :0.0000   Median :2.000   Median :1.0000  
##  Mean   :0.6745   Mean   :2.329   Mean   :0.5446  
##  3rd Qu.:1.0000   3rd Qu.:3.000   3rd Qu.:1.0000  
##  Max.   :3.0000   Max.   :3.000   Max.   :1.0000  
##  NA's   :5        NA's   :2

skim(data)

Variable type: numeric

1.2 Dataset features

This dataset is hosted on Kaggle (Heart Disease UCI), and it was from UCI Machine Learning Repository. There are records of about 300 patients from Cleveland and the features are described in a following section.

Attribute Information:

1. age
2. sex
3. chest pain type (4 values)
4. resting blood pressure
5. serum cholestoral in mg/dl
6. fbs: fasting blood sugar > 120 mg/dl

Hereon, variables are related to a nuclear stress test. That is, a stress test where a radioactive dye is also injected to the patient to see the blood flow:

7. restecg: resting electrocardiographic results (values 0,1,2)
8. thalach: maximum heart rate achieved
9. exang: exercise induced angina 10.oldpeak: ST depression induced by exercise relative to rest
10. slope: the slope of the peak exercise ST segment
11. ca: number of major vessels (0-3) colored by flourosopy
12. thal: 3 = normal; 6 = fixed defect; 7 = reversable defect
13. target: 0 = yes; 1 = No

TASK TO DO:

1. Remove X column.
2. Transform categorical variable to R factors.
3. Give (if necessary) a better name to the factor values (it will be helpful for the graphs).

# remove X column
try(data <- data %>% dplyr::select(-X))

# Transform categorical variable to R factors

data <- data %>%
    mutate(across(c(
        sex, cp, fbs, restecg, exang, slope, thal, target, ca
    ), as.factor))

# Give a better name to the factor values for the graphs
levels(data$sex) <- c("Female", "Male")
levels(data$cp) <- c("Asymptomatic", "Atypical angina", "No angina", "Typical angina")
levels(data$fbs) <- c("No", "Yes")
levels(data$restecg) <- c("Hypertrophy", "Normal", "Abnormalities")
levels(data$exang) <- c("No", "Yes")
levels(data$slope) <- c("Descending", "Flat", "Ascending")
levels(data$thal) <- c("Fixed defect", "Normal flow", "Reversible defect")
levels(data$target) <- c("Yes", "No")

Next step: Inspect all variables and make your hypotheses of how each variable affect heart attack incidence (target column).

ggplot(data, aes(target, fill=target)) + 
  geom_bar(width = .6) +
  labs(x="Disease", y="Number of patients") +
  guides(fill="none") + MY_THEME

data %>%
    ggplot(aes(x = target, y = age)) +
    geom_violin(width = .6) +
    geom_boxplot(width = .15) +
    geom_point(
        aes(col = target),
        position = position_jitterdodge(jitter.width = .3),
        alpha = .2,
        size = 3
    ) +
    theme_pubclean() +
    MY_THEME

data %>%
    ggplot(aes(x = sex, fill = target)) +
    geom_bar(width = .5) +
    MY_THEME

numeric_variables <- which(unlist(lapply(data, is.numeric)))
pheatmap(cor(data[, numeric_variables[-6]]))

ggpairs(
    data = data,
    columns = 1:4,
    aes(color = target),
    legend = 1
) +
    theme_bw()

vars <- data %>% colnames()

my.plot <- function(var){
    if (class(data[, var]) == "factor") {
        p <- data %>%
            ggplot(aes_string(x = var, fill = "target")) +
            geom_bar(width = .5, alpha = .9) +
            ggtitle(paste0(var, " vs target"))
    } else {
        p <- data %>%
            ggplot(aes_string(x = "target", y = var)) +
            geom_violin(width = .6) +
            geom_boxplot(width = .15) +
            geom_point(
                aes_string(col = "target"),
                position = position_jitterdodge(jitter.width = .3),
                alpha = .2,
                size = .5
            ) +
            ggtitle(paste0(var, " vs target")) +
            guides(col = "none")
    }
    return(p)
}

p <- lapply(vars, my.plot)

p %>% 
    wrap_plots(ncol = 3, guides = "collect") +
    guide_area() &
    MY_THEME

data %>%
    dplyr::select(c(target, restecg)) %>%
    table() %>%
    kableExtra::kable()

2.1 Random Sampling

Using the library rsample and its corresponding functions:

• initial_split
• training
• testing

Remember to use the function set.seed in order to replicate the results.

set.seed(123) #important in order to replicate

split_basico <- initial_split(data, prop = .7)
sb_train <- training(split_basico)
sb_test <- testing(split_basico)

p1 <- sb_train %>%
    ggplot(aes(x = "target", fill = target)) +
    geom_bar(position = "fill", width = .4) +
    MY_THEME

p2 <- sb_test %>%
    ggplot(aes(x = "target", fill = target)) +
    geom_bar(position = "fill", width = .4) +
    MY_THEME

wrap_plots(list(p1, p2), guides = "collect")  # se aprecia un pequeño desbalance de las clases

kableExtra::kable(table(split_basico$data$target) %>% prop.table())

kableExtra::kable(table(sb_train$target) %>% prop.table())

kableExtra::kable(table(sb_test$target) %>% prop.table())

Se aprecia que el balance de las clases del target no es constante. Vamos a estratificar para garantizar que en test y train este balance se mantenga.

2.2 Stratified Sampling

If we want to explicitly control the sampling so that our training and test sets have similar Y distributions, we can use stratified sampling.

This is more common with classification problems where the response variable may be severely imbalanced (e.g., 90% of observations with response “Yes” and 10% with response “No”).

Check the help of the function initial_split to see how to do it.

We can use the functions table and prop.table to check if training and test sets have similar Y distributions.

set.seed(123) #important in order to replicate

split_strat <- initial_split(data, prop = .7, strata = "target")
strat_train <- training(split_strat)
strat_test <- testing(split_strat)

p1 <- strat_train %>%
    ggplot(aes(x = "target", fill = target)) +
    geom_bar(position = "fill", width = .4) +
    MY_THEME

p2 <- strat_test %>%
    ggplot(aes(x = "target", fill = target)) +
    geom_bar(position = "fill", width = .4) +
    MY_THEME

wrap_plots(list(p1, p2), guides = "collect")  # clases balanceadas

kableExtra::kable(table(split_strat$data$target) %>% prop.table())

kableExtra::kable(table(strat_train$target) %>% prop.table())

kableExtra::kable(table(strat_test$target) %>% prop.table())

Ahora sí, la proporción Yes/No es constante en todo el dataset, en test y en train.

3.1 Imputation of missing values

We can use the function vis_miss of the library visdat that provides a glance ggplot of the missingness inside a dataframe.

sum(is.na(data))

## [1] 7

vis_miss(data, cluster = TRUE)

Then, to impute missing values we can use the recipe R package.

It has the following steps:

1. Function recipe: A recipe is a description of the steps to be applied to a data set in order to prepare it for data analysis.

2. step_impute_xxxx: creates a specification of a recipe step that will substitute missing values: step_impute_mean creates a specification of a recipe step that will substitute missing values of numeric variables by the training set mean of those variables. step_impute_knn creates a specification of a recipe step that will impute missing data using nearest neighbors. Can be applied to both numeric and categorical variables.

Preparamos una receta para imputar los NAs en todos los predictores.

my_recipe <- recipe(target ~ ., data = strat_train) %>% 
    step_impute_knn(all_predictors())

3. prep: For a recipe with at least one preprocessing operation, estimate the required parameters from a training set that can be later applied to other data sets.

Entrenamos los modelos KNN que se van a usar para imputar, usando los datos de train.

trained_recipe <- prep(my_recipe,training = strat_train)

4. bake: For a recipe with at least one preprocessing operation that has been trained by prep, apply the computations to new data.

Finalmente imputamos los NAs con los modelos entrenados.

datos_train_prep <- bake(trained_recipe, new_data = strat_train)
datos_test_prep <- bake(trained_recipe, new_data = strat_test)
wrap_plots(list(vis_miss(datos_train_prep, cluster = T), vis_miss(datos_test_prep)), guides = "collect", nrow = 2)

Ahora vemos que no hay ningún NA.

4.1 Logistic regression using glmnet

We are going to use the function glmnet. From its help page:

“Fit a generalized linear model via penalized maximum likelihood. The regularization path is computed for the lasso or elasticnet penalty at a grid of values for the regularization parameter lambda. Can deal with all shapes of data, including very large sparse data matrices. Fits linear, logistic and multinomial, poisson, and Cox regression models.

set.seed(123)

glmnet_traindata <- data.matrix(datos_train_prep[, -14])

fit_logistic_regression <-
    glmnet(
        x = glmnet_traindata,
        # it must be a matrix: use data.matrix
        y = datos_train_prep$target,
        family = "binomial",
        lambda = 0,
        intercept = TRUE
    )

set.seed(123)
fit_LogReg_cv_lasso <- cv.glmnet(x = glmnet_traindata,
                                 y = datos_train_prep$target,
                                 family = "binomial",
                                 nfold=10,
                                 alpha=1,
                                 type.measure = "auc")
plot(fit_LogReg_cv_lasso)

fit_LogReg_cv_ridge <- cv.glmnet(x = glmnet_traindata,
                                 y = datos_train_prep$target,
                                 family = "binomial",
                                 nfold=10,
                                 alpha=0,
                                 type.measure = "auc")
plot(fit_LogReg_cv_ridge)

fit_LogReg_cv_en <- cv.glmnet(x = glmnet_traindata,
                                 y = datos_train_prep$target,
                                 family = "binomial",
                                 nfold=10,
                                 alpha=0.2,
                                 type.measure = "auc")
plot(fit_LogReg_cv_en)

4.2 Logistic regression using caret.

control_train <-
    trainControl(
        method = "cv",
        # which type: boost, cv, none of them.. etc
        number = 10,
        # number of folds or number of resampling iterations.
        returnResamp = "all",
        classProbs = TRUE,
        search = "grid",
        savePredictions = TRUE
    )

set.seed(123)
modelo_glm_caret <- train(
    target ~ .,
    method = "glmnet",
    family = "binomial",
    trControl = control_train,
    data = datos_train_prep,
    # train data.
    metric = "Accuracy"
)

modelo_glm_caret

## glmnet 
## 
## 211 samples
##  13 predictor
##   2 classes: 'Yes', 'No' 
## 
## No pre-processing
## Resampling: Cross-Validated (10 fold) 
## Summary of sample sizes: 190, 189, 191, 191, 189, 190, ... 
## Resampling results across tuning parameters:
## 
##   alpha  lambda        Accuracy   Kappa    
##   0.10   0.0005010359  0.8208009  0.6329692
##   0.10   0.0050103594  0.8103463  0.6107406
##   0.10   0.0501035939  0.8246537  0.6419031
##   0.55   0.0005010359  0.8158009  0.6214351
##   0.55   0.0050103594  0.8151082  0.6200658
##   0.55   0.0501035939  0.7878788  0.5685067
##   1.00   0.0005010359  0.8158009  0.6214351
##   1.00   0.0050103594  0.8151082  0.6200658
##   1.00   0.0501035939  0.7917316  0.5756382
## 
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were alpha = 0.1 and lambda = 0.05010359.

plot(modelo_glm_caret) 

modelo_glm_caret$bestTune

##   alpha     lambda
## 3   0.1 0.05010359

performance <- modelo_glm_caret$resample

kbl(performance) %>%
  kable_paper() %>%
  scroll_box(width = "100%", height = "200px")

performance$alpha <- as.factor(performance$alpha)
performance$lambda <- as.factor(performance$lambda)
ggplot(data = performance, aes(x = alpha, y = Accuracy,color=lambda)) +
  geom_boxplot() +
  geom_point(position=position_jitterdodge())+
  labs(x = "") +
  theme_bw() 

control_train <-
    trainControl(
        method = "cv",
        # which type: boost, cv, none of them.. etc
        number = 10,
        # number of folds or number of resampling iterations.
        returnResamp = "all",
        classProbs = TRUE,
        summaryFunction = twoClassSummary,
        #a function to compute performance metrics across resamples.
        search = "grid",
        savePredictions = TRUE
    )

lambda <- c(0,0.01, 0.1)
alpha <- c(0,0.1,0.3,0.5, 0.9, 1)
hyper_grid <- expand.grid(alpha = alpha, lambda = lambda)

set.seed(123)
modelo_glm_caret_grid <- train(
    target ~ .,
    method = "glmnet",
    family = "binomial",
    trControl = control_train,
    data = datos_train_prep,
    tuneGrid = hyper_grid,
    metric = "ROC"
)

modelo_glm_caret_grid

## glmnet 
## 
## 211 samples
##  13 predictor
##   2 classes: 'Yes', 'No' 
## 
## No pre-processing
## Resampling: Cross-Validated (10 fold) 
## Summary of sample sizes: 190, 189, 191, 191, 189, 190, ... 
## Resampling results across tuning parameters:
## 
##   alpha  lambda  ROC        Sens       Spec     
##   0.0    0.00    0.8829461  0.7588889  0.8780303
##   0.0    0.01    0.8829461  0.7588889  0.8780303
##   0.0    0.10    0.8871886  0.7700000  0.8613636
##   0.1    0.00    0.8674411  0.7500000  0.8787879
##   0.1    0.01    0.8737542  0.7377778  0.8787879
##   0.1    0.10    0.8871801  0.7600000  0.8530303
##   0.3    0.00    0.8666077  0.7500000  0.8787879
##   0.3    0.01    0.8754966  0.7377778  0.8787879
##   0.3    0.10    0.8774579  0.7500000  0.8363636
##   0.5    0.00    0.8674411  0.7500000  0.8787879
##   0.5    0.01    0.8764057  0.7377778  0.8696970
##   0.5    0.10    0.8680640  0.7277778  0.8363636
##   0.9    0.00    0.8674411  0.7500000  0.8787879
##   0.9    0.01    0.8727189  0.7477778  0.8613636
##   0.9    0.10    0.8305724  0.7066667  0.8000000
##   1.0    0.00    0.8674411  0.7500000  0.8787879
##   1.0    0.01    0.8736448  0.7477778  0.8613636
##   1.0    0.10    0.8277189  0.6855556  0.8000000
## 
## ROC was used to select the optimal model using the largest value.
## The final values used for the model were alpha = 0 and lambda = 0.1.

plot(modelo_glm_caret_grid)

modelo_glm_caret_grid$bestTune

##   alpha lambda
## 3     0    0.1

performance <- modelo_glm_caret_grid$resample
kbl(performance) %>%
    kable_paper() %>%
    scroll_box(width = "100%", height = "200px")

performance$alpha <- as.factor(performance$alpha)
performance$lambda <- as.factor(performance$lambda)
ggplot(data = performance, aes(x = alpha, y = ROC,color=lambda)) +
    geom_boxplot() +
    geom_point(position=position_jitterdodge())+
    labs(x = "") +
    theme_bw() 

5.1 train

#for models created with glmnet
pred_fit2 <- predict(
    object = fit_logistic_regression,
    newx = data.matrix(datos_train_prep[, -14])
)

pred_fit3 <- predict(
    object = fit_LogReg_cv_lasso,
    newx = data.matrix(datos_train_prep[, -14]),
    s = "lambda.min"
)

pred_fit4 <- predict(
    object = fit_LogReg_cv_ridge,
    newx = data.matrix(datos_train_prep[, -14]),
    s = "lambda.min"
)

pred_fit5 <- predict(
    object = fit_LogReg_cv_en,
    newx = data.matrix(datos_train_prep[, -14]),
    s = "lambda.min"
)


#for models created with caret:
pred_caret <- predict(
    object = modelo_glm_caret,
    newdata =  datos_train_prep[-14],
    type = "prob"
)
pred_fit6 <- pred_caret$No  # la clase 0 en este caso es que si tenga ataque


pred_caret <- predict(
    object = modelo_glm_caret_grid,
    newdata =  datos_train_prep[-14],
    type = "prob"
)
pred_fit7 <- pred_caret$No 

labels <- (as.vector(datos_train_prep$target) == "No") + 0 #we are predicting not to have a heart attack
mis_models <- mmdata(
    list(
        as.vector(pred_fit2),
        as.vector(pred_fit3),
        as.vector(pred_fit4),
        as.vector(pred_fit5),
        as.vector(pred_fit6),
        as.vector(pred_fit7)
    ),
    labels,
    modnames = c(
        "logistic_regression",
        "lasso",
        "ridge",
        "elastic-net",
        "caret",
        "caret_grid"
    )
)
auroc <- evalmod(mis_models)
autoplot(auroc)

5.2 test

Predictions:

#for models created with glmnet
pred_fit2 <- predict(
    object = fit_logistic_regression,
    newx = data.matrix(datos_test_prep[, -14])
)

pred_fit3 <- predict(
    object = fit_LogReg_cv_lasso,
    newx = data.matrix(datos_test_prep[, -14]),
    s = "lambda.min"
)

pred_fit4 <- predict(
    object = fit_LogReg_cv_ridge,
    newx = data.matrix(datos_test_prep[, -14]),
    s = "lambda.min"
)

pred_fit5 <- predict(
    object = fit_LogReg_cv_en,
    newx = data.matrix(datos_test_prep[, -14]),
    s = "lambda.min"
)


#for models created with caret:
pred_caret <- predict(
    object = modelo_glm_caret,
    newdata =  datos_test_prep[-14],
    type = "prob"
)
pred_fit6 <- pred_caret$No  # la clase 0 en este caso es que si tenga ataque


pred_caret <- predict(
    object = modelo_glm_caret_grid,
    newdata =  datos_test_prep[-14],
    type = "prob"
)
pred_fit7 <- pred_caret$No 

Plots final:

labels <- (as.vector(datos_test_prep$target) == "No") + 0 #we are predicting not to have a heart attack
mis_models <- mmdata(
    list(
        as.vector(pred_fit2),
        as.vector(pred_fit3),
        as.vector(pred_fit4),
        as.vector(pred_fit5),
        as.vector(pred_fit6),
        as.vector(pred_fit7)
    ),
    labels,
    modnames = c(
        "logistic_regression",
        "lasso",
        "ridge",
        "elastic-net",
        "caret",
        "caret_grid"
    )
)
auroc <- evalmod(mis_models)
autoplot(auroc)